Cost of capital, returns on VC funds and risk management

1. RISKRISK
&&
RETURNSRETURNS
Prof.Stephen OngProf.Stephen Ong
BSc(Hons)Econs (LSE), MBA (Bradford)BSc(Hons)Econs (LSE), MBA (Bradford)
Visiting Professor, Shenzhen UniversityVisiting Professor, Shenzhen University
Academic Fellow, Entrepreneurship & Innovation,Academic Fellow, Entrepreneurship & Innovation,
The Lord Ashcroft International Business School,The Lord Ashcroft International Business School,
Anglia Ruskin University Cambridge UKAnglia Ruskin University Cambridge UK
MSC TECHNOPRENEURSHIP :MSC TECHNOPRENEURSHIP :
VENTURE CAPITAL FINANCINGVENTURE CAPITAL FINANCING

2. Today’s OverviewToday’s Overview

3. LEARNING OBJECTIVESLEARNING OBJECTIVES
To understanding key measures ofTo understanding key measures of
returns in the VC industry;returns in the VC industry;
To review real returns by VC firms;To review real returns by VC firms;
To understand the cost of capital forTo understand the cost of capital for
VC firms;VC firms;
To learn relevant models to includeTo learn relevant models to include
various risk factorsvarious risk factors

4. 1.Venture Capital Returns

5. Fund-level returns: Data
 Venture Economics
 Collects data from GPs, publishes vintage-year specific quartile
performance data while keeping anonymity of individual funds
 Freedom-of-Information-Act (FOIA) requests
 Forces public pensions to disclose performance of their fund
holdings
 Private Equity Performance Monitor
 Collects, packages and sells fund-specific performance data for a
fee.
 Assigns quartile rankings to funds

6. VE Benchmarks

7. Return Definitions (1)Return Definitions (1)
 Internal rate of returnInternal rate of return = a rate of return that implies an NPV of 0
for a given cash flow stream
 Value multipleValue multiple = realization ratio = investment multiple =
multiple of money = times money = absolute return
 Value multiple =
 Realized Value multiple =
 Unrealized value multiple =
feesmanagementcapitalinvested
sinvestmentunrealizedofvalueLPstoonsdistributiTotal
+
+
feesmanagementcapitalinvested
LPstoonsdistributiTotal
+
feesmanagementcapitalinvested
sinvestmentunrealizedofvalue
+

8. 5 steps to calculating value multiples
There are 3 components that contribute to value multiples: Total
(cumulative) distributions to LPs, value of unrealized
investments, and contributed capital.
1. Calculate distributions to LPs each year.
• Distributions to LPs = total distributions - carry
1. Sum them to date to get total distributions to LPs
2. Get value of unrealized investments (after exits) = portfolio value
(before exits) - total exits in year t. This is an estimate value of
illiquid investments and not a market/transaction value.
3. Calculate contributed capital = invested capital + fees to
date.
4. Calculate

9. 5 steps to calculating IRRs for LPs
As LPs, there are 3 components that contribute to your net cash
flows: fees, new investments, and distributions.
1. Calculate fees
2. Calculate distributions to LPs
• Distributions to LPs = total distributions - carry
1. Calculate net cash flows = Distributions to LPs - new investments
- fees
2. For the IRR of a fund that is T years into its life and is still alive,
the value of unrealized (i.e., remaining) investments at the end of
year T is counted as if it is a positive cash flow. This is an
estimate value of illiquid investments and not a
market/transaction value.
• Cash flow if final year of IRR calculations = Distributions to LPs -
new investments - fees + portfolio value of remaining unrealized
investments
1. IRR(year 1,…, year X) = IRR(CF1, …, CFT)

10. EXERCISEEXERCISE
IRRIRR
CALCULATIONCALCULATION

11. The IRR is not perfect
Cannot be compared to time-weighted returns
Compounding of periodic returns
Realized vs. unrealized investments
Difficult to make risk adjustments

12. Example of a J-curve

13. Carried interestCarried interest
 Contributed capitalContributed capital = invested capital + management fees that have
been paid to date
 For a fully-invested and completed fund, contributed capital = investment
capital + lifetime fees = committed capital
 Carried interest timing
 Return all call carry basis (committed or investment capital) first (25%)
 Return all contributed (or invested) capital plus priority return first (45%)
 Return only part of contributed/invested capital
 Often distinguishes between realized and unrealized investments
 Fair value test (14%)
 Other (16% of sample)

14. GP ClawbackGP Clawback
 Given the often complex formulas for distributions, GPs
could end up with more than their share of profits (excess
carry) at the end of the fund’s life.
 Clawback ensures that LPs get back what is promised to them
in the agreement by requiring GPs to return any excess carry.
 Most funds have a clawback clause.
 Many top-quartile funds raised in the late 90’s enjoyed early
carry distributions during the boom years, but then had the
clawback kick in in later years.

15. Clawback is not perfect
 Enforcing clawback is easier said than done.
 Potential disagreements between LPs and GPs over what is owed
 GPs are often not obligated to return taxed part of distributions
 Some partners may have retired or died; need to have joint and several
liability to go after the other remaining partners
Only 33% of funds make other partners liable
 Two remedies sought by GPs and LPs:
 Escrows: keep early carry distributions in escrow accounts till the end
 Annual true-ups: don’t wait till the end, re-calculate the right amount
owed each year, and seek speedy repayment.

16. Industry Returns
Industry returns are constructed as time-weighted
returns (e.g., annualized compound returns)
Nice for comparison with market indices
Nice for making risk adjustments
3 sources:
Sand Hill Econometrics (SHE): portfolio comp level
Cambridge Associates (CA): fund level
Venture Economics (VE)

17. Return DefinitionsReturn Definitions
 Periodic returnsPeriodic returns =
 Compound returnsCompound returns = (1+R1)*(1+R2)* …*(1+RT) - 1
 Annualized returnAnnualized return = = (1+ compound return)(1/T)
- 1
 Gross returns = returns before subtracting fees and carry
 Net returns = returns after subtracting fees and carry
 Realized returns = historical returns
 Expected returns = returns forecast for the future
1
)(
1
−
+
=
−t
tt
t
P
DP
R
R

18. EXERCISEEXERCISE
RETURNS :RETURNS :
Exercise 3.1 (Metrick 2011)Exercise 3.1 (Metrick 2011)

19. A Gross-Return Index

20. A Net-Return Index
Note: 1. 1981 Q1 to 2008 Q4. 2. 1988 Q4 to 2008 Q4.
13.0%1
, 16.2%2
9.0%1
, 7.9%2

21. Kleiner Perkins Returns
1
Only $170M of Fund VII was ever drawn.
Note: There have been no publicly available updates of KPCB funds since December 2004.
Fund IX’s performance as of March 2004 (-23.3% IRR) does NOT reflect its subsequent profit from
the investment it made in Google.

22. Gross vs. net value multipleGross vs. net value multiple
 GVM (gross value multiple) =
 Value multiple =
GVM is the raw investment return; value multiple is how
much money LPs makes, net of fees and carry
If GPs report their track records in terms of GVM, what
would that imply about value multiple?
capitalinvested
carrysinvestmentunrealizedofvalueLPstoonsdistributiTotal ++
feesmanagementcapitalinvested
sinvestmentunrealizedofvalueLPstoonsdistributiTotal
+
+

23. GVM, value multiple, and GP%GVM, value multiple, and GP%
 Suppose the fund is fully-invested and completed.
 Total distributions = GVM*investment capital
 Carried interest = carry%*(total distributions – carry basis)
GP% is the percentage of the total distributions that gets paid to GPs as carried
interest.
capitalCommitted
)Carry basis-capitalinvestment*(GVM*carry%-capitalinvestment*GVM
capitalCommitted
LPstoonsdistributiTotal
multipleValue ==
capitalinvestment*GVM
Carry basis)-capitalinvestment*(GVM*carry%
onsdistributiTotal
interestCarried
GP% ==

24. 2.The Cost of Capital for VC

25. Cost of capital for VCCost of capital for VC
 Historically, annualized VC return index raw return is superior to
those of public stock market indices.
 Individual investment outcomes vary greatly.
 Venture = investments with high variance in outcomes
 30-40% go bankrupt
 20-25% return 5 times or higher
 What should investors expect to earn from investing in VC?
 Not the entrepreneurs!
 Not the venture capitalists!
 What’s missing so far?

26. CAP ModelCAP Model
 Our starting point is the Capital-Asset-Pricing Model
(CAPM). It states
ri = Ri = Rf + β(Rm – Rf) ,
where
ri is the cost of capital for asset i,
Ri is the expected return for asset i,
Rf represents the risk-free rate,
Rm is the return on the whole market portfolio,
β, or “beta”, is the level of risk for asset i.
The difference (Rm – Rf) is called the market premium.

27. Risk
We make a key distinction between two kinds of risks that
are potentially present in any investments.
1. Beta risk = market risk = non-diversifiable risk =
systematic risk = “covariance”
2. Idiosyncratic risk = diversifiable risk = firm-specific risk
= residual risk = “variance”
• Why should we care which kinds of risks it is?
• Should investors demand higher returns for holding all
risks?
• Why shouldn’t all high risks be associated with higher
returns?

28. Risk (1): banana bird risk
 Consider an economy with 100 islands, 100 trees on each island, and
200 bananas per tree = 2M bananas every year.
 There lives 1 person per island, 100 total, who only consume (and care
about) bananas.
 Though they always like more bananas, their marginal utility from eating
an additional banana is decreasing in the number of bananas they eat.
 Suppose there is only one risk in this world: banana birds randomly
land on half of all islands each year and eat all bananas.
 With 50% chance, an islander gets 20,000 bananas a year.
 With 50% chance, he/she gets 0 bananas.
 Globally, birds gets 1M bananas, and people get 1M.

29. Utility with Bird Risk
 With this serious banana bird risk, their expected utility is B, weighted
average of A and D (50%*U(0) + 50%*U(20,000))
 B is worse than C (utility of getting 10,000 bananas with certainty)
 Can islanders do better?
Bananas
Utility
D
A
C
B
20K10K

30. Solution: diversificationSolution: diversification
 Each islander sets up a company holding 100 banana trees on the island, and
issues 100 shares.
 Each sells 99 shares to all others, and buys 1 share each from all others.
 There are no transaction costs.
 Now, every islander has claims on 1 banana tree on every island
 Since banana bird risk is random, every islander will get 50%*100*200 =
10,000 bananas/year with certainty. Risk is diversified away.
 Expected return on these investments = 0.
Islander’s utility after partial diversification Islander’s utility after complete diversification

31. Risk (2): Weather risk
 Now suppose that there are no banana birds, but instead
there is the following risk every year:
 50% chance of sunny year, which produces 150 bananas per tree
 50% chance of rainy year, which produces 50 bananas per tree
 With this weather risk, islander’s expected utility Y,
weighted average of X (U(5,000)) and Z (U(15,000)).
Bananas
Utility
Z
X
C
Y
15K10K5K

32. Diversifiable and non-diversifiable risk
 Weather risk affects the whole economy in a rainy year, so
the previous solution would not work.
 There is perfect covariance among all islands.
 Some islanders may agree to sell rights to their bananas in a
rainy year for rights to someone else’s bananas in a sunny
year.
 No one would give up 100 bananas in a rainy year for only 100
bananas in a sunny year.
 To give up bananas in an already rainy year, they would demand a
positive return on the deal.
 E.g., 50 bananas in a rainy year in return for 150 bananas in a
sunny year.
 This is analogous to beta risk in CAPM model.
 In contrast, with the banana bird risk, nobody would earn
extra return by agreeing to bear it, because it was free (as a
group) to get rid of it.

33. EstimatingEstimating VC cost of capitalVC cost of capital
 To estimate VC cost of capital according to CAPM model, we use
historical aggregate VC industry return data and estimate the following
equation:
Rvc, t - Rf,t = α + β(Rm,t – Rf,t) + evc,t,
where β, Rvc,t, Rm,t, and Rf,t are as defined before, except that previously
the return variables represented expected returns, while here they
represent realized (= historical) returns for period t. The new elements
in this equation are α, or “alpha”, the regression constant, and evc,t, the
regression error term.
Alpha represents the unexpected portion of the return, and positive
alpha is interpreted as skills of portfolio managers.

34. CAPM Estimation resultsCAPM Estimation results
 Two data sources
 Beta is smaller than 1 here, but CAPM is not perfect, and
we will make 3 further adjustments.
***, **, and * Indicates statistical significance at the 1, 5, and 10% level, respectively. .

35. Adjustment (1): The Fama-French Model (FFM)Adjustment (1): The Fama-French Model (FFM)
 The Fama-French (3-factor) Model has become part of a standard
tool kit for cost of capital estimation (much like CAPM) in the last
25 years.
 It is based on empirical observations that certain styles of
investments, such as “small stocks” or “value stocks” do not fit
CAPM model well.
 Now the equation is
Rvc,t- Rft= α + β * (Rmt – Rft) + βsize
* SIZEt + βvalue
* VALUEt + evc,t
where α, β, Rmt, Rft, evc,t are defined as in the CAPM, SIZEt and
VALUEtare the returns to portfolios of stocks that capture
correlations with these styles, and βsize
and βvalue
are the regression
coefficients on these returns.

36. Adjustment (2): The Pastor-Stambaugh Model (PSM)Adjustment (2): The Pastor-Stambaugh Model (PSM)
 Another relevant issue for VC is that investors may require
premium for illiquid investmentspremium for illiquid investments.
 The PSM model incorporates this illiquidity riskilliquidity risk as the
additional factor to the Fama-French 3-factor model.
 The equation is
Rvc, t- Rft= α + β * (Rmt – Rft) + βsize
* SIZEt +
βvalue
* VALUEt + βliq
* LIQt + evc,t
where LIQ is the new liquidity factor, βliq
is its regression
coefficient, and all other variables are as defined before.
 More sensitive the return on an asset is to the change in this
liquidity factor, the higher premium the investors demand
when liquidity factor return is high in the economy.

37. Adjustment (3): stale-prices
 The last issue is that aggregate VC industry price index is
updated based on reported valuations of private portfolio
companies.
 Often these are based on the most recent round of
financing, which leads to stale prices.
 We include values from past periods in our regression.
 So our final equation incorporating all 3 adjustments is
, , ,
23 11
0 0
11 11
0 0
*( ) *
* *
size
vc t ft m t s f t s s t s
value liq
s t s s t s
s
s s
s s
R R R R SIZE
VALUE LIQ
β β
β β
− − −
− −
= =
= =
− =
+ +
− +∑ ∑
∑ ∑
23
0
s
s
β β
=
= ∑
+ evc, t and

38. Final estimation results
 Market beta is now close to 2.
 Alpha is no longer significantly different from 0.
 Using both sets of estimates and taking the mid-point,
rVC= 15%
 0.04 (rf) + 1.63 * 0.07 (Market) – 0.090 * 0.025 (size) – 0.68 * 0.035 (value) + 0.26 * 0.05
(liquidity) = 14.1% (SHE estimates)
 0.04 (rf) + 2.04 * 0.07 (Market) +1.04* 0.025 (size) – 1.46 * 0.035 (value) + 0.15 * 0.05
(liquidity) = 16.6% (CA estimates)
 We use 15% as the cost of capital for VC in this course.
***, **, and * Indicates statistical significance at the 1, 5, and 10% level, respectively.

39. Firm age, capital inflow and performanceFirm age, capital inflow and performance
 If the asset class as a whole earns no alpha, whatIf the asset class as a whole earns no alpha, what
about the top VCs?about the top VCs?
Study of Venture Economics data shows that
For each VC firm, later funds do better, on
average (IRR increases with fund sequence
number)
Accumulated expertiseAccumulated expertise
Improved deal flowImproved deal flow
Better networkBetter network
New fund flows are bad for overall VC returns
(“money chasing deals”), but has no impact on
the most experienced VC

40. Does VC performance persist?Does VC performance persist?
 The study also find strong evidence that “winners stay“winners stay
winners”.winners”.
 High returns in previous funds forecast high returns in future
funds.
 This is completely different from the evidence found in
mutual fund industry, where performance does not persist in
the long run
 Implications for LPs: Access to the coveted, high-performers
is key.

41. Cost of capital for international VCCost of capital for international VC
 What cost of capital should (big,What cost of capital should (big,
diversified)diversified) investorsinvestors expect toexpect to
earn from investing inearn from investing in
international VC?international VC?
 Again, notAgain, not entrepreneursentrepreneurs
 NotNot venture capitalistsventure capitalists

42. International Portfolio DiversificationInternational Portfolio Diversification
 Because foreign stocks are not perfectly correlated with domestic
returns, combining foreign and domestic assets significantly reduce
risk.
Standard deviation of portfolio returns canStandard deviation of portfolio returns can
be reduced with internationalbe reduced with international
diversificationdiversification

43. Estimating beta with data limitationsEstimating beta with data limitations
 For companies in many countries, data needed to obtainFor companies in many countries, data needed to obtain good estimatesgood estimates
of their global beta are unavailable.of their global beta are unavailable.
 Analysts often decompose the beta to two components, one that capturesAnalysts often decompose the beta to two components, one that captures
the industry’s correlation with a national stock market, and another thatthe industry’s correlation with a national stock market, and another that
captures the nation’s correlation with the global stock market.captures the nation’s correlation with the global stock market.
 ββG,i(software)G,i(software) == ββG,IndiaG, India ** ββIndia,softwareIndia, software
 For example, for a software companyFor example, for a software company ii in India, βin India, βIndia,softwareIndia,software is how softwareis how software
industry companies are correlated with the Indian stock market, andindustry companies are correlated with the Indian stock market, and ββG,IndiaG,India
represents how Indian stock market is correlated with the global stockrepresents how Indian stock market is correlated with the global stock
market (G)market (G)
 What if βWhat if βIndia,softwareIndia,software is not available?is not available?
 ThenThen ββG,iG, i == ββG,XG,X ** ββX,iX,i ≈≈ ββG,XG,X ** ββUS,iUS,i wherewhere
 ββUSUS,i,i = estimate of beta for U.S. companies in the software industry with= estimate of beta for U.S. companies in the software industry with
respect to the U.S. stock market, used as a proxyrespect to the U.S. stock market, used as a proxy
 ββG,XG,X = estimate of country beta for country X with respect to the global= estimate of country beta for country X with respect to the global
marketmarket

44. COUNTRY BETAS, SELECTED COUNTRIESCOUNTRY BETAS, SELECTED COUNTRIES
Note: calculated based on daily index stock returns
from 1998 to 2008.
Note: calculated based on daily index stock returns from 1998 to 2008.

45. Cost of Capital for International VCCost of Capital for International VC
 We make model adjustments similar to those we made for calculating the
cost of capital for domestic (U.S.) VC.
 From Class 2 (Chapter 4): We estimated U.S. VC cost of capital as:
rrUSUS
VCVC == rrff + β+ βmarketmarket,VC,VC ** ((rrmm – r– rff)) + β+ βsizesize,VC,VC * SIZE* SIZE
+ β+ βvaluevalue,VC,VC * VALUE + β* VALUE + βliqliq,VC,VC * LIQ,* LIQ, where
rUS
VC= U.S. VC return, rf = risk-free rate,
rm – rf = market premium,
SIZE, VALUE, and LIQ are factor premia for the three factors, and
βsize,VC, βvalue,VC, and βliq,VCare their betas.
Using estimates of 4% (risk-free) 7% (market premium), 2.5% (size), 3.5%
(value), and 5% (liquidity), and using beta estimates, we obtain
rUS
VC = 15% (average of the two from Cambridge and Sand Hill data).
This gives us:
Βmarket*(rm–rf) + βsize*SIZE + βvalue*VALUE +βliq*LIQ = 15% − 4% = 11%.

46. A global multifactor model for VCA global multifactor model for VC
 Since we almost always lack data to do this exercise for non-U.S. countries,
we use the same trick using beta decomposition into country beta (βGX) and
U.S. domestic (market, size, value, liquidity, instead of industry) beta.
 Keep everything in US$.
 G = global, X= country X, US = U.S.
 For VC investment in country X, its cost of capital, rX
VC , is:
rX
VC = rf + βmarket(G),VC(X) * (RG
m – RG
f) + βsize(G),VC(X) * SIZEG
+ βvalue(G),VC(X)* VALUEG
+ βliq(G),VC(X) * LIQG
= rf + βGX* βX
market,VC * (Rm – Rf) + βGX * βX
size,VC* SIZE
+ βGX *βX
value,VC* VALUE + βGX * βX
liq,VC* LIQ
≈ rf + βGX* βUS
market,VC * (Rm – Rf) + βGX * βUS
size,VC* SIZE
+ βGX *βUS
value,VC* VALUE + βGX * βUS
liq,VC* LIQ
= rf + βGX*
[ βUS
market,VC* (Rm – Rf) + βUS
size,VC* SIZE
+ βUS
value,VC * VALUE + βUS
liq,VC* LIQ ]
= 4% + βGX* [15% − 4% ].

47. Objections and extensions
Style effects
Currency risk
Country risk
Segmented markets

48. CASE II :
YAHOO 1995

49. Further ReadingFurther Reading
 Metrick, Andrew and Yasuda, Ayako (2011) Venture
Capital & the Finance of Innovation. 2nd
Edition. John
Wiley & Sons.
 Lerner,Losh, Hardymon, Felda and Leamon, Ann
(2012). Venture Capital and Private Equity : A
Casebook. 5th
Edition. John Wiley & Sons.
 Dorf, R.C. and Byers, T.H. (2008) Technology
Ventures – From Idea to Enterprise 2nd
Edition,
McGraw Hill

50. QUESTIONS?